Average on the go

Values arrive one at a time (sensor readings, rewards, losses) and you can’t keep them all. Write update(average, count, new_value): the new average after one more value arrives, from the old average and count, how many values that average was over. At the start nothing has arrived: average and count are both 0.